Is marriage poisonous? Are relationships taxing? An analysis of the male marital wage differential in Denmark.
The word for "married" in Danish is the same as the word for "poison." The word for "sweetheart" in Danish is the same as the word for "tax." In this paper, we expand on the literature that documents a significant marital wage premium for men in the United States to see if a similar differential exists for married men in Denmark--or if the homonyms have perhaps less of a double meaning.
The existence of a marital wage premium for white men in the United States has been well documented empirically. Criticisms have focused on researchers' failure to clearly ascertain why wages change with marital status and on the imperfect nature of the data sets employed in the analysis, which generally contain relatively few men who have never married and incomplete marital histories. We use a large, 10-year panel sample of young Danish men in order to address these concerns. We have a complete relationship history for every respondent and a large fraction of never-married men. Substantial U.S.-Danish differences in marriage and childbearing behavior as well as in social norms regarding relationships and intrahousehold specialization are exploited to generate predictions regarding the Danish results that are tested in the empirical analysis. Of particular interest are the prevalence of cohabiting relationships in Denmark that allows us to test for wage differences by type of relationship, the evidence that Danish households are less specialized than U.S. households that allows us to explore the nature of the marital wage differential, and the very different pattern of childbirth and marriage that allows us to test for a distinct fatherhood effect. If wages are directly linked to productivity and if relationship type, intrahousehold specialization, and/or parenthood are linked with market wage differentials, policymakers should be apprised of the full cost of social legislation designed to alter these household choices.
2. Literature Review
The observation that married men earn more than men who have never married is not in itself surprising. Married men are typically older than never-married men, and older men have more experience, hence higher earnings, than their younger counterparts. Yet there also exists substantial evidence (for a review of the U.S. literature, see Ribar 2004) that married men earn more than never-married men with the same level of education, experience, and other observable characteristics. This fact can be explained in a number of ways.
Men who marry may be more productive throughout their lives than men who do not marry. This greater productivity makes them better providers and hence better marriage partners. This possibility can be explored econometrically either by simultaneously modeling both the decision to marry and wages (Nakosteen and Zimmer 1987; Chun and Lee 2001) or by using panel data on wages to estimate fixed-effects models that control for all unobservable, individual-specific, time-invariant factors (an early example being Korenman and Neumark 1991), or by using twins studies to control for twin-specific effects (Antonovics and Town 2004; Krashinsky 2004). Results indicate that there are differences between men who marry and men who do not. Korenman and Neumark (1991) conclude that 20% of the marital wage differential is attributable to individual-specific and time-invariant factors. Gray (1997) reports similar results using a cohort of men born in 1942-1952 but finds that for younger cohorts in the United States (born 1958-1965), all the estimated marital differential is attributable to fixed effects. Krashinsky (2004) finds that controlling for twin-specific effects explains the entire differential, but Antonovics and Town (2004) find that twin-based controls for selection yield even larger marital wage differentials.
The idea that marriage may change a man's productivity has also received some attention in the literature. One theoretical explanation is drawn from Becker (1991) and based on the fact that individuals in joint households are more able to specialize than those in single-person households. Men have historically specialized more in the market sector and women more in the home sector. This leaves men more time and/or energy to spend on market work after marriage. If this translates to higher productivity on the job, then their earnings should immediately rise. In this case, the level of wages will rise as men marry but fall back down if/ when the marriage ends. The marital wage effect will be temporary. Alternatively, men who marry may specialize by increasing their investment in job-related human capital. In this case, married men's wages may not rise immediately but will rise more rapidly, and wage growth--but not necessarily wage level--will fall if/when a marriage ends.
There is indirect empirical evidence from selectivity-controlled estimates supporting both these specialization mechanisms in the United States. Some researchers have found evidence that wages do rise more rapidly for married men (Korenman and Neumark 1991; Gray 1997 for older U.S. cohorts born 1942-1952; Stratton 2002), and some have found that wages both jump and rise faster following marriage (Daniel 1991; Hersch and Stratton 2000). However, this evidence is indirect because it does not actually capture behavioral changes in effort or time use.
Few data sets provide direct measures of productivity or time use. Evidence that married men receive more training than unmarried men is provided in Rodgers and Stratton (2005) but not found to explain the marital wage differential. Mehay and Bowman (2005) provide direct evidence of labor force productivity differentials between married and unmarried men but do not examine wages. A number of researchers have inferred that intrahousehold specialization will vary inversely with the employment status/hours of the wife and so compared marital wage differentials for men with employed wives to those for men whose wives are not employed (Loh 1996; Hotchkiss and Moore 1999). Results are mixed, with Loh finding men married to more educated wives faring better in the labor market and Hotchkiss and Moore finding results that differ depending on the husband's occupation. More direct evidence on men's housework activities suggests that in the United States, while men's wages are negatively related to their housework time, controlling for men's time on housework does not explain the marital wage differential (Hersch and Stratton 2000).
Other explanations for a male marital wage differential include discrimination, marriage as a behavior-altering state that focuses men on more productive activities, and a compensating wage differential argument that suggests that married men favor income over other job characteristics (for a more detailed summary, see Ribar 2004). Parenthood also may generate effects if becoming a parent changes men's behavior on the job. Most marital wage researchers control for the presence of children in the household and fail to find a significant impact (see, e.g., Korenman and Neumark 1991; Loh 1996). Mehay and Bowman (2005) find mixed empirical results but conclude that marital duration has an impact on performance that is independent of the presence of children. One exception is Cornwell and Rupert (1997), who find that fathers earn about 5% more than nonfathers in the United States and hypothesize that, like married men, fathers modify their time allocation decisions in a way that increases their market productivity. Generally, however, in the United States, it may be difficult to distinguish between marriage and fatherhood, as the latter so often follows fairly closely after the former.
There are a number of problems with both the evidence and the theory behind the marital wage differential as presented to date. One concern with the selection hypothesis is that virtually all men eventually marry. In the United States, 63% of all white, non-Hispanic women are married by age 25, 81% by age 30 (Bramlett and Mosher 2002). Those who never marry are but a small and likely unusual fraction of the population.
Not only is this a problem with the hypothesis, but it also poses problems empirically, as a marital wage differential can be identified only by comparing married and not-married individuals. Samples including persons of all ages are unlikely to include many never-married men. Estimates based on youth cohorts have a better chance of including more first-time marriages, but even these samples include a substantial fraction of men who are married when first observed (78% in Korenman and Neumark's 1991 seminal work, 76.2% in Gray 1997, 66.2% in Hersch and Stratton 2000). In part this is due to sample selection criteria that restrict the sample to men who have completed their education, but the result is that, in general, estimates of the marital wage differential rely a great deal not on first marriages but on separation/divorce and remarriage for identification of the marital wage premium (for a further discussion, see Cornwell and Rupert 1997).
In addition, much of the literature ignores cohabitation. Exceptions include Schoeni (1995), who finds no effect of cohabitation on earnings in Germany; Loh (1996), Stratton (2002), and Bardasi and Taylor (2004), who conclude that any effect of cohabitation is transitory; and Cohen (2002) and Richardson (2003), who find that cohabiting men receive a smaller premium than married men. The work by Stratton (2002) is of particular interest as it employs a data set that has both panel data and cohabitation histories. Controlling for individual-specific effects, Stratton finds that married men but not cohabiting men earn significantly more than men not in a relationship. Given the trend toward declining marriage and increasing cohabitation rates over the past 30 years (Bumpass, Sweet, and Cherlin 1991), the effect of cohabitation on men's wages warrants further attention.
3. What Can Danish Data Tell Us?
What contribution can an analysis of Danish data bring to this literature? Data from the United States have provided some evidence of a white male marital wage differential, but as data on actual job productivity are typically unavailable, the nature of the differential is more often inferred than positively identified. International data may provide some additional insights, as there are substantial cross-national differences in the timing and type of interpersonal relationships, in the household division of labor, and in the timing of paternity.
There already exists evidence that married men earn a premium in many developed nations (Schoeni 1995). However, there also exists evidence that the nature of the premium differs. Both Richardson (2003), using Swedish data, and Bardasi and Taylor (2004), using British data, report that while wages jump following marriage, they do not rise faster following marriage. Indeed, Richardson finds some evidence that in Sweden wage growth is lower for married than for unmarried men. Other work by Ginther, Sundstrom, and Bjorklund (2006) looking at Swedish parents finds that the entire marital wage effect is attributable to selection. Denmark is much more similar to its Scandinavian neighbor Sweden than to the United States as regards the nature of interpersonal relationships, the household division of labor, and social welfare policy. Thus, there is reason to expect that the Danish male marital wage differential will differ from that observed in the United States. No such evidence is yet available. Naur and Smith (1998), in an analysis of gender wage differentials, provide some evidence that there is a Danish male marriage premium (of approximately 4%), but while their work differentiates between married and cohabiting men, it does not control for the possibility of differential wage growth.
Differences in interpersonal relationships between Denmark and the United States arise both in their timing and in their type. Men marry later in Denmark than in the United States. Looking at men aged 25-29, 50.8% in the United States have been married compared to 18.1% in Denmark. Even at age 30-34, 70.4% of U.S. men but only 46.3% of Danish men report having been married. Overall, marriage appears to be a more selective state in Denmark than in the United States, and thus the selection effect of marriage may be greater in Denmark. (1)
One explanation why men in Denmark marry less often and later is the greater prevalence of cohabitation. While couples in the United States typically cohabit for only a short spell and often do so as a prelude to marriage or following a divorce (Forste 2002), cohabitation is a much more socially acceptable and enduring relationship in Denmark. In 2001, 22% of all couple households in Denmark were cohabiting (Statistics Denmark 2001), as compared with only about 6% in the United States (Fields and Casper 2001). Looking at first cohabitations in the United States, 21% of these couples split up, and 30% marry within one year, while 10% split up and 3% marry within one year in Denmark. At the end of three years, 39% have split up, and 35% have married in the United States, while 38% have split up and 15% have married in Denmark. The wage effects of cohabitation are likely to be greater in Denmark than in the United States because the longer-lasting cohabitations in Denmark are likely to be both more selective and to have more impact on labor market productivity than cohabitations in the United States.
As our analysis here focuses exclusively on Danish data, it is important to generate predictions for the impact of relationship type in Denmark. While cohabiting relationships in Denmark are more enduring than cohabiting relationships in the United States, cohabiting relationships are less enduring than marriages in Denmark. While 2% of first marriages end within one year and 13% within three years, the comparable figures for first cohabitations are 10% and 39%. Legally, there are likely fewer differences between married and cohabiting couples in Denmark than in the United States, but even in Denmark there are substantial distinctions, particularly during the period captured by our sample. (2) For example, cohabiting persons are treated like single persons in determining inheritance, public income transfers, and social benefits. This means that in some cases there is an economic incentive to marry (income taxes) and in others there is not (transfer payments). In the case of housing subsidies, there is no distinction between married and cohabiting persons, as it is household, not individual, income that determines these benefits. As in the United States, it is easier to terminate a cohabiting relationship than a marriage. As regards child custody following the dissolution of a relationship, joint custody is the default for married couples, while the mother is given preference in cohabiting households. Generally, despite the greater prevalence of cohabitation in Denmark versus the United States, we still expect marriage to be a more selective state than cohabitation.
The difference in the relative stability of marriages and cohabitations also suggests there will be differences in intrahousehold specialization. Couples in less stable relationships will likely engage in less intrahousehold specialization. This is particularly likely if there is any cost associated with specialization. To the extent that the relationship wage differentials are the result of a productivity change associated with intrahousehold specialization, we predict that the productivity change associated with cohabiting relationships will be smaller than the productivity change associated with marriage. Thus, any jump in wages or wage growth should be smaller for cohabiting than for married men.
The different nature of relationships in Denmark generates some wage predictions; evidence of cross-national differences in intrahousehold specialization leads to others. There is some evidence that the degree of specialization even in married households may be lower in Denmark than in the United States. In the 1990s, Denmark (joint with Sweden) had the highest female labor force participation rate in the OECD (see Jaumotte 2003). Further, women entering the labor market during the 1980s and 1990s were typically working full time, not part time like earlier generations. Recent evidence shows that Danish women, unlike women in most other countries (Smith et al. 2003), do not reduce their labor supply significantly when they become mothers, except during maternity leave. By way of contrast, the labor force participation rate for married women in the United States with children under the age of three was only 58.0% in 2002, while it was 80.5% for married women with children aged 14-17 (Statistical Abstract of the United States 2003). One reason why Danish households appear to engage in less labor market/nonlabor market specialization is that the Danish public sector has taken over a large part of the care work for children, the sick, and the elderly.
International comparisons based on time use surveys also show that Danes (jointly with Swedes) are less likely to specialize in household production activities (Bonke and Koch-Weser 2003). In Table 1, the development of the U.S. and Danish gender distribution of housework is shown. In 1965 there appeared to be more specialization within Danish households. In that year, Danish men contributed only 10% as much time toward housework as women, while in the United States men contributed about 30%. In both countries, the gender division of household labor has become more equal over time. But this change has been more dramatic in Denmark. By 1985, intrahousehold specialization, as measured by the ratio of male-to-female housework hours, was equal between Denmark and the United States. By 2003, Danish households appeared to be less specialized than U.S. households. Furthermore, these changes have entailed much more significant behavioral changes on the part of Danish men. On average, men in the United States reported spending the same amount of time on housework in 1985 and 2003, while men in Denmark increased their housework time by half. With less intrahousehold specialization, Danish men may experience a smaller increase in productivity following marriage than U.S. men. (3) This would suggest a smaller jump and/or a smaller change in the growth rate of earnings following marriage for Danish as compared to U.S. men.
Finally, while there is little evidence that fatherhood influences wages in the United States, it may be the case that becoming a father is a more significant turning point in the lives of Danish men than either marriage or cohabitation. One hypothesis might be that specialization within the household really starts or changes when the spouse (mother) enters her first maternal leave period following childbirth. While any maternity leave available in the United States is so short lived that it would make little sense to reallocate household tasks, in Denmark there is an almost 100% coverage of publicly funded maternal leave, which since 1984 has had a duration of 26 weeks and since 2002 a duration of one year. This leave introduces a major "shock" to Danish families. Whether this shock introduces a temporary or a permanent change in behavior remains to be seen, but fatherhood rather than relationship type may be more closely linked with wage changes in Denmark. As with marriage, there are also significant differences between the United States and Denmark regarding the timing of fatherhood and marriage. They are not nearly as closely linked in Denmark as in the United States. Thus, it may also be easier to identify the effect of fatherhood on men's wages in Denmark than in the United States.
In this study, we use a large panel sample of young Danish men. All our data come from official Danish registers. These are governmental records akin to U.S. Social Security or Internal Revenue Service-based data matched to local marriage and school records. These data include information on earnings and employment as well as marital and parental status for all residents of Denmark. Our initial sample consists of a 10% sample of the Danish-born population of men born between 1966 and 1975 inclusive--a total of 37,881 men aged 18 or younger in 1984. Register data on these individuals is available annually from 1984 until 2001, yielding 564,788 observations on men aged 16-35.
Marriage histories are obtained directly from official administrative records that record precise dates of marriage and divorce and so provide much more accurate data than respondent surveys, which are dependent on personal recall. Studies of the marital wage differential typically treat separated and divorced couples similarly, as specialization is contingent on having a partner in residence. We follow this protocol and classify married men whose spouse is not present as separated if the spouse does not return in the following year. Men who reside with an unrelated female who is within 15 years of age are classified as cohabiting. This is the definition employed by Statistics Denmark. Coresident partners are each assigned a unique identifier that enables us to determine whether the partner changes from year to year, but no further partner information is available. Because coresidence information is available only annually, cohabiting relationships that last less than one year will be undersampled. Such relationships are, however, unlikely to have any substantial influence on earnings, so the loss is relatively minor. In addition, as individuals are not directly asked the nature of their relationship with nonmarital partners, cohabitation is measured with some error. Roommates may be incorrectly classified as cohabiters, and individuals who have substantially older partners or cohabiting partners who maintain a separate legal address may be incorrectly classified as noncohabiters. Comparison of register and survey data from another Danish source (the 2001 Danish Time Use Survey) suggests that while more cohabiting persons are misclassified as single than vice versa, the margin of error is very small. In general, the Statistics Denmark classification is deemed accurate.
As the data allow us to distinguish between different partners, we can observe partner changes and identify cohabiters who go on to marry. As we observe individuals from a very young age, we are able to construct a very comprehensive history of both marriages and cohabitations for every respondent. This is in contrast to Richardson (2003), who has information only on the duration of the current marriage, no complete marital history, and no information on how long any cohabiting couple has been together, and to Stratton (2002), who has only incomplete cohabitation records. Restricting the sample to those not missing information on marital status or history, to those not presently widowed, to those never observed in a gay union, and to those who are not fathers when first observed reduces our sample size by less than 100 men to 548,054 observations on 37,802 individuals.
Employment data are also obtained from the Danish register. These official records report earnings, education, job experience, occupation, and industry. In order to keep as comprehensive a sample as possible, we do not restrict our wage analysis to those who have completed their education. However, in order to analyze wages, we do restrict our analysis to individuals for whom we have information on education, occupation, industry, and labor market experience; who have wage reports of between 40 and 800 Danish kroner (DKr) per hour (between approximately 5 and 96 U.S. dollars); who worked more than 20% of the year (about 320 hours); and who are not self-employed. The hours and self-employment restrictions are necessary in order to reliably construct hourly wage information. Further, all individuals who are observed only once are excluded, as they contribute nothing to panel estimates. This leaves us with a primary sample of 297,938 observations on 33,798 individuals. Nonemployment accounts for almost half the observations lost at this stage, lack of a reasonable wage measure accounts for another quarter, and the rest are dropped because they are observed only once. In order to more nearly match restrictions imposed in the U.S.-based literature, a second sample that excludes those who have not completed their education and those who worked less than 80% of the year (about 1280 hours) is constructed. As about 18% of our sample is still enrolled at age 25, this sample is substantially smaller, containing 172,883 observations on 24,951 individuals.
Sample statistics for selected variables in the larger sample of 33,798 individuals are reported in Table 2 for both the 2001 cross section and the pooled panel data set covering the period 1984-2001. In the Appendix, sample statistics for all variables included in the analysis are shown. According to Table 2, as of 2001, 34% of the 25,548 men in the sample were married, 34% cohabiting, and among those men not observed in a relationship, 37% had previously been cohabiting and 8% married. Thus, even in the last year of the panel, cohabitation is clearly more widespread than legal marriage. In 2001, married men earned on average 127 DKr hourly ($15.27 2001 U.S.), compared to only 117 DKr ($14.06) and 112 DKr ($13.46) for cohabiting and single men. These raw statistics indicate that married men earned 13% more and cohabiting men 4% more than men not currently in a relationship. Part of this raw marital wage differential is certainly attributable to the fact that the married men were, on average, about two years older than the nonmarried men.
Table 3 presents statistics pertaining to changing marital status, to fatherhood, and to educational enrollment status. At the age of 18, almost none of the 33,798 individuals who are included in this study were married (0%), and only 2% were cohabiting. By age 35, 53% were legally married, and 24% were cohabiting. More than 2% are in a relationship when they first appear in our estimation sample because to enter our sample they must be employed, and some do not begin work until a later age. However, even when we restrict the analysis to those in our estimation sample who have completed school, only 5% are married and 27% cohabitating when first observed. As a result, we are able to achieve a cleaner identification of the marriage and cohabitation wage differentials than was possible in previous studies where a majority of the sample entered married. A comparison of the fraction of married men and of fathers demonstrates what we claimed earlier--that fatherhood does not closely follow marriage in Denmark. Indeed, fatherhood seems to precede marriage in Denmark, as the fraction who are fathers at any age always exceeds the fraction who are married. Finally, Table 3 demonstrates the prevalence of late enrollment in Denmark. Fifty-one percent of the men were enrolled in school at the age of 18, and 10% were enrolled at age 27. Fortunately, almost all the men had left the educational system by age 35. (4) Of course, since we restrict this sample to individuals with "reliable" wage information who were employed for at least 20% of the year, Table 3 is not representative of all young Danish men. A substantial number of full-time students who do not have a job are excluded.
The analysis is based on a traditional human capital wage function:
in [W.sub.it] = [[beta].sub.t] + [X.sub.it] [beta] + [Z.sub.it] [gamma] + ([alpha].sub.i] + [[epsilon].sub.it), (1)
where [X.sub.it] is a vector of explanatory variables and [Z.sub.it] is a vector of family status variables. The subscripts i and t index the individual and time, respectively; [[alpha].sub.i] is the unobserved heterogeneity term, assumed to be individual specific, time invariant; and [[epsilon].sub.it] is an error component, Nid(O, [[sigma].sup.2.sub.[[epsilon]). The explanatory variables include five industry dummies, eight occupation dummies that take into account changes in occupational designations beginning in 1996, two region dummies, four education dummies, and quadratic measures of actual labor market experience. The vector of family status variables, [Z.sub.it], includes indicators for mutually exclusive marriage, cohabitation, divorce/separation, and past cohabitation; quadratic duration measures for these states; and variables reflecting the age and presence of children and of fatherhood. While we report primarily on estimates of 7 in what follows, explanatory variables X are included in every specification, and full model results are available on request.
Robust standard errors corrected for clustering at the individual level are reported for all the pooled ordinary least squares (OLS) models. If, however, the individual specific component, [[alpha].sub.i], is correlated with any element of X or Z, then OLS estimates of [beta] and [gamma] will be biased. The particular concern here is that men who marry may be more productive (have a higher [alpha]) than men who do not marry. In this case, the dummy variable identifying married men will be positively correlated with [alpha], and the impact of marriage on wages will likely be overstated. Nevertheless, we begin by reporting both OLS and selection-corrected results in order to gauge the importance of the selection correction.
We replicate the standard marital wage equations in which only dummy variables identifying men who are currently married or are not now married but have been in the past (divorced) are included in Z. Table 4, columns 1 and 2, shows pooled OLS and fixed-effects estimates for the sample of young Danes working full time who have completed school (sample 1). A small but significant marriage premium of about 2% is found in the pooled OLS regressions, but when controlling for unobserved individual effects, that is, selectivity into marriage, the marital wage premium seems to disappear. Only for those individuals who divorce are wages significantly different (lower) than for those individuals who have never been legally married. These results indicate a much lower marital wage premium than typically found in U.S. studies.
However, these findings are in part due to the specification of the model. In columns 1-2, the excluded category contains both single and cohabiting men. In columns 3-4, we treat married and cohabiting men the same, comparing all partnered men and all men not currently partnered who have been in the past (divorced or previously cohabited) to all never partnered men. This treatment causes the estimated fixed-effects relationship wage premium to become significantly positive though small--about 1% for both currently married/cohabiting men and men who are not now in a relationship but have married or cohabited in the past. Finally, columns 5-6 show the estimated coefficients from a specification in which marriage and cohabitation are treated as distinct relationships. In this specification, the type of current relationship is identified and, for those not currently in a relationship, the most recent type of relationship. As found in Stratton (2002), the cohabitation premium is smaller than the marriage premium (p-values for a test of equality are 0.000 in both specifications). Pooled OLS estimates show a marriage premium of 4.2% (almost double that observed when not controlling for cohabitation status) and a cohabitation premium of 2.5%. These estimates are smaller than those observed in the United States, which are often on the order of 8-14%, but about half the differential is explained by the smaller wage dispersion in Denmark. Figures presented in Datta et al. (2006) indicate that a Danish wage differential of 4.2% is roughly equivalent to a 6.5% wage differential in the United States. The remaining differential may be attributable to the fact that Danish households specialize less than U.S. households. As predicted, the selectivity corrected premia are substantially smaller than the OLS premia, indicating that selectivity effects are quite important for this group of young men working full time.
Sample 1 is a sample of young men working full time who have left the educational system either without a formal education or with a completed education. As a large number of Danish men in the age-group 20-30 years are still enrolled in school, this selection mechanism tends to oversample low-skilled, uneducated men. Although we control for education and occupation, the estimated marriage premia found in columns 1-6 may be misleading because of potential unobserved (time varying) factors. (5) In columns 7-12, the same models are estimated using a sample of all young men for whom we observe reliable wage information (sample 2). This sample includes students who work while enrolled in school, a situation that is very common among Danish students. (6) Extending the sample clearly has a positive effect on the estimated wage premia from marriage and especially cohabitation in all specifications. The marriage premium in the final specification (column 12) is 1.6%, 30% higher than sample 1 estimates, and of the same magnitude for cohabiters, indicating that married and cohabiting men earn the same wage premium--as long as they stay with their partners! If they separate, those who were cohabiting experience no wage loss, while those who were legally married see their wages fall back to the level of men who have never married. A comparison with column 11 indicates that selection accounts for about two-thirds of the marital premium and 50% of the cohabitation premium. As hypothesized, the selection effect is substantial and larger for married than for cohabiting men.
The experience accumulated in jobs as students may have a different effect on wages than experience accumulated in jobs after school completion. If some men marry only after completing their education, we may mix the effects from marriage and cohabitation with effects from differential wage growth following completion of school and education. Therefore, to allow the model to be more flexible with respect to the effect of experience, we split the experience variable into experience accumulated before the highest observed level of education was completed and experience accumulated after completing education. Selected coefficients from this estimation are shown in Table 5. As expected, the wage effect of accumulated experience from "student jobs" is smaller than the wage effect of experience accumulated after completing education. The more flexible specification of the wage function matters for the relative size of the estimated wage premia for legally married and cohabiting men. The premium decreases from 1.6% to 1.4% for legally married men, while it increases from 1.6% to 2.0% for cohabiting men. For those who divorce and for cohabiting couples that split, the difference between a legal marriage and cohabitation seems even more pronounced. Men who divorce earn about as much as men who have never been in a relationship, while men who were formerly cohabiting earn about 2% more. This differential may be related to the nature of the intrahousehold specialization effect. Married men may experience a boost in immediate productivity, while cohabiting men invest more time in enhancing future productivity. To explore this possibility, it is necessary to control for time married and cohabiting as well as the state itself.
In Table 6, we extend the preferred flexible wage model that includes separate pre- and posteducation experience measures with measures of the duration of different marital and cohabitation states (column 2) and, additionally, their square terms (column 3). Taken together, these results indicate that there is little evidence of intrahousehold specialization benefiting married men in the long run. Marriage does lead to an immediate increase in earnings, which may be attributable to specialization that immediately increases time and/or energy on the job. However, this wage boost at least partially disappears when the marriage ends. Furthermore, wages do not rise faster for married men but, as in Sweden (Richardson 2003), rise more slowly. (7) We hypothesized that, because there appears to be less intrahousehold specialization in Denmark than in the United States, there would be less or perhaps no differential wage growth following marriage. Our results go beyond the predictions of theory, perhaps suggesting that marriage reduces Danish men's investment in market-related human capital. This contrasts with the results for cohabiting men who experience an immediate and permanent boost in wages and for whom there is some evidence that wages rise faster but at a diminishing rate. It is this difference in wage growth during the relationship between cohabiting and married men that drove the previously observed postcohabitation premium. At the same time, wage growth is even slower after couples break up than it is during the relationship itself. As these results control for selection, they suggest that breaking up and returning to the single life may reduce productivity on the job--though this effect ceases with the commencement of the next relationship.
Alternative specifications (results available on request) were also tested. Specifications controlling only for time with the current partner yield similar results, and wage growth patterns appear to be similar for past and current relationships. When we restrict the sample to only never-married men, our estimates of the impact of cohabitation on both the level and the growth rate of wages are as reported in the paper. When we restrict the sample to men in a relationship, we find very little difference in the level of wages for married versus cohabiting men and a slower rate of wage growth both during and following marriage as compared to cohabitation. These results are consistent with those reported for the full sample. Overall, sensitivity testing suggests that the results reported here are robust.
These results provide evidence of a small but statistically significant relationship wage premium in Denmark for young men. This premium is somewhat smaller than the marital premium found in U.S. studies and is negatively, not positively, associated with the duration of the relationship. As expected, we find that selection has a smaller impact on cohabitation than on marriage, but, unlike Stratton (2002), we find that cohabiting men do receive a wage premium even after controlling for selection. This is likely due to the different, more enduring cohabitations observed in Denmark that do encourage some behavioral changes. (8)
To pursue this hypothesis further, we finally analyze whether the small observed marital and cohabitation wage premium of about 1.5-2% is really a premium to having a partner or whether it may instead reflect a childbirth-related premium captured because of the relation between marriage/cohabitation and childbirth. The relation between marriage and childbirth is less pronounced in Denmark than in the United States and so is more easily estimated in Denmark. For simplicity, we drop the quadratic duration terms as we add fatherhood-related variables. In Table 7, column 2, we include an additional variable to identify households with children aged 0-17 years. Apparently, the presence of a child has no significant effect on the estimated relationship wage premium. However, when adding more detailed information on the age of the children, we find that the age matters. The presence of a child less than the age of three increases men's wages, but when the child passes the age of three, the positive initial effect becomes a significant negative effect. Furthermore, the effect of marriage and cohabitation seems to decrease slightly when controlling for the presence of children in different age categories.
Paralleling our controls for relationship duration, in our final specification we include a variable measuring the duration of fatherhood. (9) Including this variable reduces the estimated relationship premia further, especially for legally married men, but the premia remain statistically significant at between 1.2% and 1.5%, and wages do not appear to drop back down when the relationship ends, suggesting that marriage and cohabitation are associated with a permanent jump in productivity. This jump is not attributable to specialization that permits individuals to spend more time/energy on the job only while in a relationship, as such an effect would end when the relationship and such specialization ends. Nor is the change associated with specialization that results in increased investment in job-related human capital, as that would be reflected in higher wage growth during marriage. There is no evidence of any wage growth differential. Indeed, the negative impact of relationships on wage growth observed in previous specifications appears to be attributable to fatherhood status rather than to the relationship itself. Only postrelationship wage growth remains significantly different from wage growth for those never married, and that is negative, potentially indicating reduced investment. These results are robust to an alternative specification in which marriage and cohabitation are not distinguished from one another (results available on request). Formal tests, however, reject this specification in favor of one that allows marriage and cohabitation to have distinct wage effects. The results regarding marital and cohabitation status in Table 7 are also quite robust to specifications including interaction terms between marital/cohabitation status and the child-related variables (results available on request). Relationship duration has no relation to earnings, but wages decline with time postrelationship. The interaction terms do indicate, however, that while married men experience a fatherhood penalty, it may be somewhat smaller than that experienced by cohabiting men.
As suggested earlier, having a child has a positive initial effect on wages that disappears within one to two years. This result may indicate that fatherhood makes men more responsible and productive at their job but only temporarily, that Danish employers consciously or unconsciously reward new fathers, or that fathers in Denmark bear substantial child care responsibilities following the period of extensive maternal leave offered by the Danish social welfare system. The fact that there is an initial boost in earnings following the first childbirth leads us to believe that the latter explanation may be the most accurate, as the period of maternal leave may provide the father with an opportunity to spend more time investing in the market.
If the father has this opportunity with the first birth, however, there is reason to suspect that he may also have this opportunity with later births, and if childbirth and marriage or cohabitation are correlated, such a "birth effect" may be driving the observed jump in wages at the onset of a relationship. To explore this possibility, we examined the frequency with which marriages and cohabitations begin in the same year as a child is born. We find that 20% of all marriages occur in the same year as the man first becomes a father and that another 7% of all marriages occur in the same year as another child is born. Childbirth and cohabitation are far less likely to occur in the same year, with only 8.5% of cohabitations occurring in the same year as a first childbirth and less than 1% occurring in the same year as a subsequent childbirth. Dropping those men who begin a relationship in the same year as they become a father so as to avoid any confusion of a relationship effect and a childbirth effect, we are able to replicate the results reported in Table 6 (results available on request). Our results regarding marriage and cohabitation in Denmark are not attributable to the generous maternity leave policies found there.
To round out our analysis, we reestimated the final specification from Table 7 with and without controls for cohabitation and for fixed effects (results available on request) in order to determine if the results we reported earlier in the paper hold as well for this, our preferred specification. As before, we find evidence that failure to control for cohabitation (at least in Denmark) substantially biases OLS estimates of the marital wage premium for men--reducing the effect by over half. Also as before, we find that selection explains just over half the cohabitation effect and over two-thirds of the marital effect on wages. This supports our hypothesis that marriage is a more selective state than cohabitation. We also examined the data for evidence that it is perhaps not only wage level but also wage growth that predicts selection into marriage. There does not appear to be clear support for such a selection mechanism. Using a set of men not in a relationship before age 22, we find on average lower wage growth between the ages of 21 and 22 for those who marry as compared to those who remain unattached in any of the next three years. Using a set of men not in a relationship before age 23, we find a slightly higher wage growth between ages 21 and 23 for those who marry in the next year or two than for those who do not marry, but the differences are not overwhelming.
In this study, we use data from Denmark to shed light on the nature of the male marital wage differential. We begin by presenting evidence that differences between Denmark and the United States in interpersonal relationships, in the intrahousehold allocation of time, and in parental responsibilities generate some predictions regarding relationship wage differentials in Denmark. Specifically, we predict that while marriage is still likely to be a more selective state than cohabitation in Denmark, the net relationship differential and in particular the component not attributable to selection is likely to be smaller in Denmark than in the United States because of the more egalitarian division of time in Denmark. We also posit that fatherhood may have a greater impact on Danish men's earnings than either marriage or cohabitation because of the generous maternity leave offered in Denmark that may promote specialization. We then introduce a panel data sample of Danish men that overcomes several data-related problems common with marital wage studies. Specifically, our data consist of a large sample of men so young that few enter the sample with experience in a relationship and for whom we have complete marital histories, substantially complete cohabitation records, and full wage information for up to 18 years from a reliable government-based data source. Thus, our results are not as subject as those from previous studies to missing data or errors-in-variables bias or to the criticism that the marriage effect is derived primarily from the effects of divorce or not adequately distinguished from cohabitation.
Overall, we find substantial support for our hypotheses. OLS estimates indicate that there is a relationship premium of between 3.2% and 4.0% in Denmark. This premium, as predicted, is smaller than that observed in the United States. Relationship type is important in that failure to control for cohabitation reduces the estimated marital wage differential by over half in Denmark because cohabiting men do share some of the wage benefits enjoyed by married men. Thus, U.S. researchers should be forewarned that controlling for cohabitation may become important in the United States as cohabitation there becomes increasingly common. When the OLS differential is split into selection and nonselection components, we find, consistent with our expectations, that while selection into cohabitation is important, it constitutes a smaller fraction of the cohabiting as compared to the marital wage differential in Denmark.
After controlling for selection, there remains a small positive and significant "relationship wage premium" of about 1.5%. Given the evidence that Danish households specialize less than U.S. households, the larger marital wage premium observed in the United States may be attributable to increased productivity as a result of greater intrahousehold specialization. When controlling for fatherhood, this relationship premium appears to consist of a one-time jump in wages. Such a jump may be attributable to specialization that increases productivity immediately, or it may be attributable to behavioral changes that arise with the advent of a serious relationship. It is difficult to disentangle these effects, but the fact that wages do not fall far or at all when the relationship ends suggests more of a permanent behavioral effect. More importantly, we do not find evidence of faster wage growth for Danish men in a relationship. An explanation consistent with these findings is that Danish men who enter a relationship do not invest additional time in market-related human capital, perhaps because, with the more egalitarian division of household tasks in Denmark, they do not have much additional time. Taken together, these results provide indirect evidence that relationship types and household choices (particularly regarding time use) influence market-based productivity. Thus, policy directives that influence household decisions in these areas may have consequences in the marketplace for men. However, while encouraging intrahousehold specialization may increase men's productivity in the marketplace, it may have an adverse effect on women. Further research on the consequences for women is necessary before making any policy recommendations.
Of perhaps greater interest are our findings regarding fatherhood. Our results show that Danish men receive a "fatherhood" premium during their first few years as fathers but that this premium is rapidly consumed by lower postfatherhood wage growth. Since we control for (time constant) selection effects by estimating fixed-effects models, we suspect that the initial jump in wages may be attributable to the increased intrahousehold specialization made possible by Denmark's generous maternity leave policy. This leave may temporarily change behavior. However, a comparison of direct child care time by Danish parents indicates that fathers contributed 0.65 hours per hour spent by mothers in 2001 (Danish 2001 Time Use Survey), while Bianchi, Wight, and Raley (2005), using figures from the American Time Use Survey 2003, reported that U.S. fathers contributed even less time and only half the time spent by mothers. Thus, there is evidence that, overall, Danish households specialize less in child care as well as household production. The observation that wage growth is lower for Danish fathers than nonfathers suggests that Danish fathers may be diverting time from on-the-job training toward child care. More information on the duration of maternity leave taken and the time use patterns of Danish men before and after childbirth is needed to explore the fatherhood effect further, but our findings show that while relationships may not be "taxing" for men in Denmark, fatherhood may be.
Sample Statistics All Years, Pooled Sample 1985 Age Age All 18 35 All Age 25.344 18 35 18.408 Hourly wage rate, DKr, 1984 prices 96.392 61.081 128.293 61.651 [X.sub.it] variables Rural area 0.169 0.227 0.185 0.214 Small city 0.564 0.615 0.582 0.615 No education 0.314 0.976 0.178 0.850 High school 0.125 0.016 0.063 0.105 Short postsecondary 0.474 0.008 0.563 0.045 Medium postsecondary (bachelor's level) 0.056 0 0.114 0 Long postsecondary (master's level) 0.031 0 0.082 0 Still enrolled 0.147 0.506 0.021 0.411 Years of actual employment experience 6.122 1.043 13.915 1.405 Years of experience posteducation 3.594 0.089 9.805 0.126 Raw materials 0.048 0.141 0.025 0.140 Manufacturing 0.273 0.325 0.282 0.343 Construction 0.122 0.113 0.116 0.101 Service 0.402 0.319 0.418 0.298 Other industry 0.010 0.006 0.014 0.006 Occupation controls (a) Other occupation 0.098 0.423 0 0.382 Salaried, medium or high level 0.011 0.001 0 0.001 Salaried, low level 0.120 0.130 0 0.144 Skilled 0.135 0.228 0 0.220 Unskilled 0.115 0.211 0 0.242 Other 1996-classification 0.101 0 0.189 0 Upper salaried, 1996 classification 0.059 0 0.183 0 Lower salaried, 1996 classification 0.275 0 0.453 0 [Z.sub.it] variables Child aged 0-17 years 0.210 0.001 0.661 0.005 Child aged 0-2 years 0.150 0.001 0.278 0.003 Child aged 3-9 years 0.109 0 0.532 0.001 Child aged 10-17 years 0.015 0 0.160 0.001 Years since individual became a father, conditional on being a father 1.010 0.001 5.784 0.005 Married 0.153 0.001 0.535 0.002 Cohabiting 0.297 0.015 0.231 0.037 Divorced or separated 0.009 0 0.044 0 Cohabited in the past 0.091 0.001 0.094 0.003 Years married 0.543 0 3.643 0.001 Years cohabited 1.661 0.009 4.630 0.025 Years divorced or separated 0.035 0 0.269 0 Years postcohabitation 0.505 0.001 1.473 0.002 No. of observations 297,938 8576 2959 3211 No. of individuals 33,798 8576 2959 3211 2001 All Married Cohabiting Age 30.698 31.900 30.071 Hourly wage rate, DKr, 1984 prices 118.620 126.624 116.804 [X.sub.it] variables Rural area 0.161 0.197 0.157 Small city 0.545 0.603 0.533 No education 0.188 0.166 0.175 High school 0.076 0.056 0.074 Short postsecondary 0.550 0.564 0.571 Medium postsecondary (bachelor's level) 0.111 0.120 0.113 Long postsecondary (master's level) 0.075 0.094 0.067 Still enrolled 0.059 0.032 0.065 Years of actual employment experience 9.930 11.504 9.428 Years of experience posteducation 7.003 8.336 6.565 Raw materials 0.027 0.027 0.026 Manufacturing 0.259 0.262 0.264 Construction 0.121 0.124 0.127 Service 0.438 0.429 0.439 Other industry 0.013 0.013 0.011 Occupation controls (a) Other occupation 0 0 0 Salaried, medium or high level 0 0 0 Salaried, low level 0 0 0 Skilled 0 0 0 Unskilled 0 0 0 Other 1996-classification 0.199 0.180 0.188 Upper salaried, 1996 classification 0.158 0.190 0.150 Lower salaried, 1996 classification 0.483 0.452 0.501 [Z.sub.it] variables Child aged 0-17 years 0.445 0.844 0.449 Child aged 0-2 years 0.276 0.511 0.303 Child aged 3-9 years 0.283 0.589 0.231 Child aged 10-17 years 0.055 0.107 0.051 Years since individual became a father, conditional on being a father 2.741 4.928 2.137 Married 0.341 1 0 Cohabiting 0.337 0 1 Divorced or separated 0.025 0 0 Cohabited in the past 0.120 0 0 Years married 1.590 4.269 0.013 Years cohabited 3.349 3.972 4.795 Years divorced or separated 0.104 0.052 0.080 Years postcohabitation 1.103 0.660 1.124 No. of observations 25,548 8707 8598 No. of individuals 25,548 8707 8598 2001 Not in a Relationship Age 30.081 Hourly wage rate, DKr, 1984 prices 112.060 [X.sub.it] variables Rural area 0.126 Small city 0.499 No education 0.228 High school 0.098 Short postsecondary 0.513 Medium postsecondary (bachelor's level) 0.099 Long postsecondary (master's level) 0.062 Still enrolled 0.083 Years of actual employment experience 8.791 Years of experience posteducation 6.052 Raw materials 0.029 Manufacturing 0.250 Construction 0.110 Service 0.448 Other industry 0.016 Occupation controls (a) Other occupation 0 Salaried, medium or high level 0 Salaried, low level 0 Skilled 0 Unskilled 0 Other 1996-classification 0.231 Upper salaried, 1996 classification 0.132 Lower salaried, 1996 classification 0.498 [Z.sub.it] variables Child aged 0-17 years 0.018 Child aged 0-2 years 0.002 Child aged 3-9 years 0.013 Child aged 10-17 years 0.004 Years since individual became a father, conditional on being a father 1.061 Married 0 Cohabiting 0 Divorced or separated 0.077 Cohabited in the past 0.371 Years married 0.284 Years cohabited 1.184 Years divorced or separated 0.184 Years postcohabitation 1.548 No. of observations 8243 No. of individuals 8243 (a) Occupational classifications changed in 1996. The First five occupations listed are those used prior to 1996. The final three are those used beginning in 1996.
We are grateful to Mette Kornvig, Astrid Wurtz, and Camilla Osterballe Pedersen for very helpful research assistance. We gratefully acknowledge financial support from the Danish Social Research Council, FSE. Leslie Stratton also gratefully acknowledges support from a 2005 Summer Research Grant from the Virginia Commonwealth University School of Business. Note that these data, like U.S. Social Security records, are not publicly available because of their detailed and sensitive nature.
Received June 2005; accepted November 2006.
Antonovics, Kate, and Robert Town. 2004. Are all the good men married? Uncovering the sources of the marital wage premium. American Economic Review 94:317-21.
Bardasi, Elena, and Mark Taylor. 2004. Investigating the marriage wage premium for men in Britain. Paper presented at ESPE meetings, Bergen, Norway.
Becket, Gary S. 1991. A treatise on the family. Enlarged edition. Cambridge, MA: Harvard University Press.
Bianchi, Suzanne, Vanessa Wight, and Sara Raley. 2005. Maternal employment and family caregiving: Rethinking time with children in the ATUS. Paper presented at the ATUS Early Results meetings, North Bethesda, MD.
Bonke, Jens, and Elke Koch-Weser. 2003. The welfare state and time allocation in Sweden, Denmark, France and Italy. In Changing life patterns in Western industrial societies, edited by Janet Zollinger Giele and Elke Hoist. New York: Elsevier, JAI, pp. 231-53.
Bramlett, Matthew D., and William D. Mosher. 2002. Cohabitation, marriage, divorce, and remarriage in the United States. Vital Health Statistics 23(22). Washington, DC: National Center for Health Statistics.
Bumpass, Larry L., James A. Sweet, and Andrew Cherlin. 1991. The role of cohabitation in declining rates of marriage. Journal of Marriage and the Family 53:913-27.
Chun, Hyunbae, and Injae Lee. 2001. Why do married men earn more: Productivity or marriage selection? Economic Inquiry 39:307-19.
Cohen, Philip N. 2002. Cohabitation and declining marriage premium for men. Work and Occupations 29:346-63.
Cornwell, Christopher, and Peter Rupert. 1997. Unobserved individual effects, marriage, and the earnings of young men. Economic Inquiry 35:285-94.
Daniel, Kermit. 1991. Does marriage make men more productive? Unpublished paper, University of Chicago. Datta Gupta, Nabanita, Ronald L. Oaxaca, and Nina Smith. 2006. Swimming upstream, floating downstream: Comparing women's relative wage positions in the U.S. and Denmark. Industrial and Labor Relations Review 59:243-66.
Datta Gupta, Nabanita, and Nina Smith. 2002. Children and career interruptions: The family gap in Denmark. Economica 69(276):609-29.
Fields, Jason, and Lynne Casper. 2001. America's families and living arrangements: March 2000. Current Population Reports, P20-537. Washington. DC: U.S. Census Bureau.
Forste, Renata. 2002. Prelude to marriage or alternative to marriage? A social demographic look at cohabitation in the U.S. Journal of Law and Family Studies 4:91-104.
Ginther, Donna K., Marianne Sundstrom, and Anders Bjorklund. 2006. Selection or specialization: The impact of legal marriage on adult earnings in Sweden. Paper presented at the EALE Conference, Prague.
Gray, Jeffrey S. 1997. The fall in men's return to marriage: Declining productivity effects or changing selection? Journal of Human Resources 32:481-504.
Hersch, Joni, and Leslie S. Stratton. 2000. Household specialization and the male marriage wage premium. Industrial and Labor Relations Review 54:78-94.
Hotchkiss, Julie L., and Robert E. Moore. 1999. On the evidence of a working spouse penalty in the managerial labor market. Industrial and Labor Relations Review 52:410-23.
Jaumotte, Florence. 2003. Female labour force participation: Past trends and main determinants in OECD countries. OECD Economics Department Working Papers No. 376.
Korenman, Sanders, and David Neumark. 1991. Does marriage really make men more productive? Journal of Human Resources 26:282-307.
Krashinsky, Harry A. 2004. Do marital status and computer usage really change the wage structure? Journal of Human Resources 39:774-91.
Lausten, Mette, and Karen Sjorup. 2003. How do men and women spend their time? Report 03:08. Copenhagen: Danish National Social Research Institute. (in Danish)
Loh, Eng Seng. 1996. Productivity differences and the marriage wage premium for white males. Journal of Human Resources 31:566-89.
Mehay, Stephen L., and William R. Bowman. 2005. Marital status and productivity: Evidence from personnel data. Southern Economic Journal 72:63-77.
Nakosteen, Robert A., and Michael A. Zimmer. 1987. Marital status and earnings of young men. Journal of Human Resources 22:248-68.
Naur, Miche1e, and Nina Smith. 1998. Cohort effects on the gender wage gap in Denmark. In Women's work and wages, edited by Inga Persson and Christina Jonung. New York: Routledge, pp. 122-44.
Ribar, David. 2004. What do social scientists know about the benefits of marriage? A review of quantitative methodologies. IZA Discussion Paper No. 998.
Richardson, Katarina. 2003. The evolution of the marriage premium in the Swedish labor market 1968-1991. Unpublished paper, Institute for Labour Market Policy Evaluation, Uppsala, Sweden.
Rodgers, William M. III, and Leslie S. Stratton. 2005. The male marital wage differential: Race, training, and fixed effects. IZA Discussion Paper No. 1745.
Schettkat, Ronald. 2003. Differences in US-German time-allocation: Why do Americans work longer hours than Germans? IZA Discussion Paper No. 697.
Schoeni, Robert F. 1995. Marital status and earnings in developed countries. Journal of Population Economics 8:351-9.
Smith, Nina, Shirley Dex, Tim Callan, and Jan D. Vlasblom. 2003. Taxation of spouses: A cross country study of the effects on married women's labour supply. Oxford Economic Papers 55:417-39.
Statistical Abstract of the United States. 2003. "Section 12: Labor Force, Employment, and Earnings." Accessed 13 August 2007. Available http://www.census.gov/prod/2004pubs/O3statab/labor.pdf.
Statistics Denmark. 2001. Statistical 10-year survey (2001). Copenhagen: Statistics Denmark.
Stratton, Leslie S. 2002. Examining the wage differential for married and cohabiting men. Economic Inquiry 40:199-212.
(1) All U.S. figures on first marriage and first cohabitation come from Bramlett and Mosher (2002). These statistics come from the 1995 wave of the U.S. National Survey of Family Growth. Danish marriage statistics come from Statistics Denmark (http://www.statistikbanken.dk) and are based on 2002 statistics. Danish cohabitation statistics come from our own sample.
(2) More recently, Denmark has been moving in a direction similar to Sweden and reducing the legal distinctions between cohabiting and married couples.
(3) One explanation for this observed difference in time allocation between Danish and U.S. men may be the large difference in tax pressure between the two countries. The overall tax pressure in Denmark has increased during the latest decades to a level of about 50%, while the U.S. level is about 30%. Schettkatt (2003) shows that a considerable part of the difference between time allocation in the United States and Germany can be explained by the tax wedge in the two countries. Since the tax wedge is larger in Denmark compared to Germany, this result is expected to carry over to Denmark.
(4) We assume that individuals who are observed not enrolled at the age of 26 have completed their education. This assumption is necessary because the youngest men in our sample are age 26 when last observed in 2001. Evidence from our oldest birth cohorts indicates that very few individuals reenroll after the age of 26.
(5) As an alternative, we might have modeled the selection into sample 1 explicitly. However, it is difficult to identify this selection because we would have to find variables that affect the educational decision but not the wage level. Therefore, we prefer to include all observations but use a flexible wage specification, which allows the experience profile to change when education is completed.
(6) We do not include an indicator for part-time work in the estimations shown because part-time work in Denmark is typically paid according to the same pay scales as full-time work. So, controlling for education and occupation, part-time work is not found to have a negative effect on hourly wage rates (see, e.g., Datta Gupta and Smith 2002).
(7) Controlling for quadratics in pre- and posteducation experience as well as time married clearly risks introducing multicollinearity. However, results (available on request) show that neither the magnitude nor the statistical significance of the experience measures changes substantially when controls for relationship duration are added to the specification.
(8) We find smaller premiums than those reported for Sweden, where Richardson (2003) found a marital wage premium of 8.5% and a cohabitation premium of about half that size. One reason for this differential may be that we analyze much younger birth cohorts of men than Richardson and that, for these cohorts, the changing role of women and families (less specialization and more market work for women) is much more pronounced than in older birth cohorts. Richardson finds evidence of declining male marital wage premiums in Sweden during the last three decades that could support this hypothesis.
(9) Coefficient estimates and standard errors to the labor market experience variables change very little with the addition of this variable. Collinearity with experience is not a problem.
Nabanita Datta Gupta, * Nina Smith, ([dagger]) and Leslie S. Stratton ([double dagger]])
* CIM, IZA, Danish National Institute of Social Research, Herluf Trollesgade 11, DK 1052 Copenhagen K, Denmark; E-mail email@example.com.
([dagger]) CIM, IZA, and Department of Economics, Aarhus School of Business, Prismet, Silkeborgvej 2, DK-8000 Aarhus C, Denmark; E-mail firstname.lastname@example.org.
([double dagger]) CIM, IZA, Virginia Commonwealth University, P.O. Box 844000, Richmond, VA 23284-4000, USA; E-mail Isstratt@vcu.edu; corresponding author.
Table 1. Average Weekly Hours Spent on Housework for Men and Women in Denmark and the United States (a) 1965 1975 Denmark Men 3 8 Women 31 27 Ratio (male/female 0.10 0.30 hours) United States Men 12 14 Women 40 33 Ratio (male/female 0.30 0.42 hours) 1985 2001-2003 Denmark Men 11 17 Women 21 24 Ratio (male/female 0.52 0.71 hours) United States Men 16 16 Women 31 28 Ratio (male/female 0.52 0.57 hours) Source: Data for Denmark are from Lausten and Sjorup (2003). U.S. data for 1965, 1975, and 1985 stem from ISR Panel Income Dynamics (http://www.umich.edu/news/? Releases/Mar02/chrO31202a). Data from 2003 are calculated from the new time use surveys from the Bureau of Labor Statistics (http://www.bls.gov/news.release/atus.txt). (a) For Denmark, data for the years shown are based on the time use studies collected in 1964, 1975, 1987, and 2001. For both countries, the data from 2001 to 2003 are based on time use data collected by other institutions or based on slightly different principles, and therefore the absolute level may not be comparable to the previous years (see Lausten and Sjorup 2003). However, the ratio of male to female hours is expected to be robust and comparable across years. Age-group for Denmark is 16-74 years, for the United States 15+. Table 2. Sample Statistics All Years 2001 All All Married Age 25.344 30.698 31.900 Hourly wage rate, DKr, 1984 prices 96.392 118.620 126.624 Married 0.153 0.341 1 Cohabiting 0.297 0.337 0 Divorced or separated 0.009 0.025 0 Cohabitated in the past 0.091 0.120 0 Child aged 0-17 years 0.210 0.445 0.844 Years married 0.543 1.590 4.269 Years cohabiting 1.661 3.349 3.972 Years since individual became a father 1.010 2.741 4.928 No. of observations 297,938 25,548 8707 No. of individuals 33,798 25,548 8707 2001 Cohabiting Not in a Relationship Age 30.071 30.081 Hourly wage rate, DKr, 1984 prices 116.804 112.060 Married 0 0 Cohabiting 1 0 Divorced or separated 0 0.077 Cohabitated in the past 0 0.371 Child aged 0-17 years 0.449 0.018 Years married 0.128 0.284 Years cohabiting 4.795 1.184 Years since individual became a father 2.137 1.061 No. of observations 8598 8243 No. of individuals 8598 8243 Table 3. Proportion Who Are Married, Cohabiting, Fathers, or Still Enrolled in School: Sample of 33,798 Individuals, Including Students with a Part-Time Job Proportion Who Are: Age Married Married or Fathers Enrolled-- Cohabiting Still in School 18 0.00 0.02 0.00 0.51 19 0.00 0.05 0.00 0.43 20 0.00 0.11 0.01 0.36 21 0.01 0.19 0.02 0.23 22 0.02 0.28 0.04 0.15 23 0.03 0.36 0.06 0.13 24 0.06 0.43 0.10 0.12 25 0.09 0.49 0.14 0.11 26 0.13 0.55 0.19 0.10 27 0.17 0.59 0.25 0.08 28 0.23 0.63 0.32 0.07 29 0.29 0.66 0.39 0.05 30 0.36 0.69 0.45 0.05 31 0.40 0.71 0.51 0.04 32 0.43 0.72 0.55 0.03 33 0.47 0.74 0.60 0.03 34 0.50 0.75 0.63 0.02 35 0.53 0.77 0.66 0.02 First observation 0.01 0.09 0.01 Table 4. Estimation of Male Hourly Wage Function for Full-Time Workers, Excluding Students (Sample 1) and All Young People with Observed Wages (Sample 2): Selected Coefficients (Standard Deviations in Parentheses) Sample 1 Full-Time Workers, Excluding Students (a) (1) (2) (3) Pooled OLS FE Pooled OLS Married 0.022 *** 0.002 (0.003) (0.002) Cohabiting Married or 0.030 *** cohabiting (0.002) Divorced -0.000 -0.016 *** (0.003) (0.005) Previously cohabited Divorced or previously 0.029 *** cohabited (0.003) No. of individuals 24,951 24,951 24,951 No. of observations 172,883 172,883 172,883 [R.sup.2] (FE: overall) 0.327 0.262 0.328 Sample 1 Full-Time Workers, Excluding Students (a) (4) (5) (6) FE Pooled OLS FE Married 0.041 *** 0.012 *** (0.003) (0.002) Cohabiting 0.025 *** 0.009 *** (0.002) (0.002) Married or 0.009 *** cohabiting (0.002) Divorced 0.020 * -0.000 (0.007) (0.005) Previously 0.031 *** 0.010 *** cohabited -0.004 -0.003 Divorced or previously 0.008 *** cohabited (0.002) No. of individuals 24,951 24,951 24,951 No. of observations 172,883 172,883 172,883 [R.sup.2] (FE: overall) 0.262 0.329 0.263 Sample 2 All Individuals with Reliable Wage Information, Including Students (a, b) (7) (8) (9) Pooled OLS FE Pooled OLS Married 0.023 *** 0.001 (0.002) Cohabiting Married or 0.037 *** cohabiting (0.002) Divorced 0.007 -0.025 *** (0.005) Previously cohabited Divorced or previously 0.039 *** cohabited (0.003) No. of individuals 33,798 33,798 33,798 No. of observations 297,938 297,938 297,938 [R.sup.2] (FE: overall) 0.452 0.404 0.453 Sample 2 All Individuals with Reliable Wage Information, Including Students (a, b) (10) (11) (12) FE Pooled OLS FE Married 0.047 *** 0.016 *** (0.003) (0.002) Cohabiting 0.033 *** 0.016 *** (0.002) (0.002) Married or 0.016 *** cohabiting (0.002) Divorced 0.031 *** -0.006 (0.007) (0.005) Previously 0.041 *** 0.016 *** cohabited -0.003 -0.002 Divorced or previously 0.014 *** cohabited (0.002) No. of individuals 33,798 33,798 33,798 No. of observations 297,938 297,938 297,938 [R.sup.2] (FE: overall) 0.405 0.453 0.405 OLS = ordinary least squares; FE = fixed effects. (a) All models include controls for region, educational level, total employment experience (and its square), occupational status, sector, and year. (b) A dummy for enrollment status is also included. * Indicates significance at the 5% level. ** Indicates significance at the 1% level. *** Indicates significance at the 0.1% level. Table 5. Estimation of Male Hourly Wage Function, Fixed-Effects Estimations of Models on Different Samples and with Different Specification of Experience: Selected Coefficients (Standard Deviations in Parentheses) (a) Sample 1 Sample 2 Sample 2 All Individuals All Individuals Full-Time with Reliable with Reliable Workers, Wage Wage Excluding Information, Information, Students Including Including Students Students Married 0.012 *** 0.016 *** 0.014 *** (0.002) (0.002) (0.002) Cohabiting 0.009 *** 0.016 *** 0.020 *** (0.002) (0.002) (0.002) Divorced -0.000 -0.006 -0.009 (0.005) (0.005) (0.005) Cohabited in the past 0.010 *** 0.016 *** 0.021 *** (0.003) (0.002) (0.002) Total employment 0.090 *** 0.069 *** -- experience (0.001) (0.001) Total employment -0.288 *** -0.228 *** -- experience, (0.004) (0.003) squared/100 Employment experience -- -- 0.044 *** before completing (0.001) education Employment experience -- -- -0.150 *** before completing (0.006) education, squared/100 Employment experience -- -- 0.058 *** after education (0.001) Employment experience -- -- -0.215 *** after education, (0.004) squared/100 No. of individuals 24,951 33,798 33,798 No. of observations 172,883 297,938 297,938 [R.sup.2] (overall) 0.263 0.405 0.403 (a) All models include controls for region, educational level, enrollment status, occupational status, sector, and year. * Indicates significance at the 5% level. ** Indicates significance at the 1% level. *** Indicates significance at the 0.1% level. Table 6. Fixed-Effects Estimation of Male Hourly Wage Function, Including Years of Marriage, Cohabitation, and Divorce: Selected Coefficients (Standard Deviations in Parentheses) Sample 2 All Individuals with Reliable Wage Information, Including Students (a) Married 0.014 *** 0.020 *** 0.016 *** (0.002) (0.002) (0.003) Cohabiting 0.020 *** 0.019 *** 0.014 *** (0.002) (0.002) (0.002) Divorced -0.009 0.012 * 0.012 (0.005) (0.006) (0.007) Cohabited in the past 0.021 *** 0.021 *** 0.015 *** (0.002) (0.002) (0.002) Years married -- -0.004 *** -0.006 *** (0.001) (0.001) Years married, squared/100 -- -- 0.016 (0.012) Years cohabiting -- -0.002 *** 0.002 * (0.000) (0.001) Years cohabiting, squared/100 -- -- -0.043 *** (0.008) Years divorced or separated -- -0.011 *** -0.016 *** (0.002) (0.004) Years divorced, squared/100 -- -- 0.085 (0.056) Years postcohabitation -- -0.003 *** 0.001 (0.001) (0.001) Years postcohabitation, -- -- -0.057 *** squared/100 (0.014) No. of individuals 33,798 33,798 33,798 No. of observations 297,938 297,938 297,938 [R.sup.2] (overall) 0.403 0.403 0.404 Test statistic and p-value for 18.72 16.28 24.44 test that cohabitation and 0.000 0.000 0.000 marriage have the same impact on earnings (including quadratic terms in column 3) (a) All models include controls for region, educational level, enrollment status, employment experience prior to completing and following completion of education (and square terms), occupational status, sector, and year. * Indicates significance at the 5% level. ** Indicates significance at the 1% level. *** Indicates significance at the 0.1% level. Table 7. Estimation of Male Hourly Wage Function, Including Variables for Children and Fatherhood: Selected Coefficients (Standard Deviations in Parentheses) Model 3 Fixed Effects: All Model Individuals with Reliable Wage Information, Including Students (a) Married 0.020 *** 0.018 *** (0.002) (0.003) Cohabiting 0.019 *** 0.018 *** (0.002) (0.002) Divorced 0.012 0.012 * (0.007) (0.006) Cohabiting in the past 0.021 *** 0.021 *** (0.002) (0.002) Years married -0.004 *** -0.005 *** (0.001) (0.001) Years cohabiting -0.002 *** -0.002 *** (0.000) (0.000) Years divorced or separated -0.011 *** -0.011 *** (0.002) (0.002) Years postcohabitation -0.003 *** -0.003 *** (0.001) (0.001) Years in fatherhood -- -- Child aged 0-17 -- 0.003 (0.002) Child aged 0-2 -- -- Child aged 3-9 -- -- Child aged 10-17 -- -- No. of individuals 33,798 33,798 No. of observations 297,938 297,938 R2 (overall) 0.403 0.404 Test statistic and p-value for 16.28 16.02 test that cohabitation and 0.000 0.000 marriage have the same impact on earnings Model 3 Fixed Effects: All Model Individuals with Reliable Wage Information, Including Students (a) Married 0.016 *** 0.012 *** (0.003) (0.003) Cohabiting 0.018 *** 0.015 *** (0.002) (0.002) Divorced 0.008 0.011 (0.006) (0.006) Cohabiting in the past 0.020 *** 0.020 *** (0.002) (0.002) Years married -0.003 *** 0.000 (0.001) (0.001) Years cohabiting -0.002 *** 0.001 (0.000) (0.000) Years divorced or separated -0.011 *** -0.007 *** (0.002) (0.002) Years postcohabitation -0.003 *** -0.002 ** (0.001) (0.001) Years in fatherhood -- -0.007 *** (0.000) Child aged 0-17 -- 0.009 *** (0.002) Child aged 0-2 0.009 *** -- (0.002) Child aged 3-9 -0.010 *** -- (0.002) Child aged 10-17 -0.012 *** -- (0.004) No. of individuals 33,798 33,798 No. of observations 297,938 297,938 R2 (overall) 0.404 0.404 Test statistic and p-value for 12.76 3.42 test that cohabitation and 0.000 0.008 marriage have the same impact on earnings (a) All models include controls for region, educational level, enrollment status, employment experience prior to completing and following completion of education (and square terms), occupational status, sector, and year. * Indicates significance at the 5% level. ** Indicates significance at the 1% level. *** Indicates significance at the 0.1% level.
|Printer friendly Cite/link Email Feedback|
|Comment:||Is marriage poisonous? Are relationships taxing? An analysis of the male marital wage differential in Denmark.|
|Author:||Gupta, Nabanita Datta; Smith, Nina; Stratton, Leslie S.|
|Publication:||Southern Economic Journal|
|Article Type:||Author abstract|
|Date:||Oct 1, 2007|
|Previous Article:||Is more information always better? An experimental study of charitable giving and Hurricane Katrina.|
|Next Article:||Negative expected value suits in a signaling model.|